Multi-parameter collaborative calibration method and device for UAV power system test platform

Abstract

This invention relates to the technical field of power system testing, specifically including a multi-parameter collaborative calibration method and apparatus for a UAV power system testing platform. The method includes: setting a calibration response path, combining priority parameters to obtain multi-parameter collaborative calibration optimization scores under various operating conditions, performing multi-objective optimization balancing of the calibration strategy, obtaining multi-parameter collaborative calibration instructions, and executing synchronous calibration and value correction of various test parameters during UAV power system testing. This solves the technical problem that the testing platform performs isolated calibration of multiple power system parameters, failing to fully consider the correlation between parameters under typical flight conditions, and thus failing to effectively suppress systematic deviations. It achieves the technical effect of constructing a multi-parameter calibration condition map, setting a calibration response path adapted to typical flight states, performing multi-objective optimization balancing of the calibration strategy to generate executable collaborative calibration instructions, and improving the consistency of multi-parameter collaboration while maintaining calibration accuracy.

Description

Technical Field

[0001] This invention relates to the technical field of power system testing, specifically to a multi-parameter collaborative calibration method and apparatus for a UAV power system testing platform. Background Technology

[0002] Unmanned aerial vehicles (UAVs) are widely used in critical fields such as logistics and emergency rescue. The performance consistency and reliability of the power system directly determine flight safety. It should be noted that UAV power systems are characterized by strong nonlinearity, multi-physics coupling, and rapid dynamic changes in operating conditions. Single parameter calibration is difficult to reflect the interactive effects in real flight. Current test platforms still adopt a discrete and sequential calibration mode, with each sensor calibrated independently. The calibration process is mostly based on static or simplified operating conditions, lacking mapping and correlation with real flight states such as hovering, climbing, and rapid descent. At the same time, calibration strategies usually focus on a single indicator, ignoring multi-dimensional requirements such as the efficiency of value reproducibility, airworthiness compliance, and calibration cycle coverage. This results in poor generalization ability of calibration results in actual flight, and may even mask systematic deviations, making it difficult to support the full life cycle quality control of highly reliable UAV systems.

[0003] In summary, existing technologies suffer from the technical problem that test platforms perform isolated calibrations of multiple parameters of the power system without fully considering the correlation between parameters under typical flight conditions, thus failing to effectively suppress systematic deviations. Summary of the Invention

[0004] This application provides a multi-parameter collaborative calibration method and apparatus for a test platform for unmanned aerial vehicle (UAV) power systems. It aims to solve the technical problem that existing test platforms perform isolated calibration of multiple parameters of the power system, without fully considering the correlation between parameters under typical flight conditions, and thus cannot effectively suppress systematic deviations.

[0005] In view of the above problems, the technical solution to achieve the present application is as follows: In a first aspect, this application provides a multi-parameter collaborative calibration method for a test platform of an unmanned aerial vehicle (UAV) power system. The method includes: screening a multi-parameter calibration dataset to identify abnormal test bench datasets requiring calibration intervention; drafting a multi-parameter calibration condition map containing parameter coupling disturbance topology and flight condition simulation distribution; setting calibration response paths adapted to typical UAV flight conditions in the multi-parameter calibration condition map; obtaining multi-parameter collaborative calibration optimization scores for each condition segment by combining priority parameters; performing multi-objective optimization balancing of the calibration strategy based on the calibration response paths and the multi-parameter collaborative calibration optimization scores for each condition segment to obtain multi-parameter collaborative calibration instructions; and using the multi-parameter collaborative calibration instructions to perform synchronous calibration and value correction of various test parameters during the UAV power system testing process.

[0006] In a second aspect, this application provides a multi-parameter collaborative calibration device for a UAV power system test platform. The device comprises: a data filtering module for filtering multi-parameter calibration data sets, identifying abnormal test bench datasets requiring calibration intervention, and drafting a multi-parameter calibration condition map including parameter coupling disturbance topology and flight condition simulation distribution; a calibration response path setting module for setting calibration response paths adapted to typical UAV flight conditions in the multi-parameter calibration condition map, and obtaining multi-parameter collaborative calibration optimization scores for each condition segment based on priority parameters; an optimization balancing module for performing multi-objective optimization balancing of calibration strategies based on the calibration response paths and the multi-parameter collaborative calibration optimization scores for each condition segment, resulting in multi-parameter collaborative calibration instructions; and a synchronous calibration module for using the multi-parameter collaborative calibration instructions to perform synchronous calibration and value correction of various test parameters during UAV power system testing.

[0007] In summary, one or more technical solutions provided in this application achieve the technical effect of constructing a multi-parameter calibration condition map, setting a calibration response path adapted to typical flight conditions, performing multi-objective optimization and balancing of calibration strategies to generate executable collaborative calibration instructions, and improving the collaborative consistency of multiple parameters while taking into account calibration accuracy. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0009] Figure 1 This application provides a flowchart illustrating the multi-parameter collaborative calibration method for an unmanned aerial vehicle (UAV) power system test platform.

[0010] Figure 2 This application provides a schematic diagram of the structure of a multi-parameter collaborative calibration device for an unmanned aerial vehicle (UAV) power system test platform.

[0011] Explanation of reference numerals in the attached diagram: Data filtering module M100, calibration response path setting module M200, optimization balance module M300, and synchronous calibration module M400. Detailed Implementation

[0012] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0013] Example 1: The present application will be described in detail below with reference to the accompanying drawings, as follows... Figure 1 As shown, this application provides a multi-parameter collaborative calibration method for a UAV power system test platform, wherein the method includes: S1: Filter the multi-parameter calibration data set to determine the abnormal test bench dataset that requires calibration intervention, and formulate a multi-parameter calibration condition map that includes parameter coupling disturbance topology and flight condition simulation distribution; S2: Set a calibration response path adapted to the typical flight condition of the UAV in the multi-parameter calibration condition map, and obtain the multi-parameter collaborative calibration optimization score under each condition segment by combining the priority parameters.

[0014] Specifically, the multi-parameter calibration dataset refers to a structured dataset formed after preprocessing the raw monitoring data collected by multiple sensor channels on the UAV power system test platform. Its data dimensions include metadata such as timestamps, device identifiers, channel identifiers, values, and operating mode codes. Among them, the multiple sensor channels correspond to speed encoders, current sensors, and temperature sensors. The parameter coupling disturbance topology is a graph structure model that describes the nonlinear interaction relationship between various test parameters of the power system. The nodes represent physical parameters such as speed, current, and temperature, and the edges represent the coupling strength and disturbance propagation path between parameters. The edge weights are usually determined by physical mechanism models or data-driven methods.

[0015] The flight condition simulation distribution is a spatial probability distribution of flight conditions constructed based on typical flight profiles for UAV airworthiness certification. It expands discrete flight condition nodes into continuous dynamic transition processes, reflecting the temporal characteristics and statistical laws of parameter changes in real flight. The typical flight profiles include hovering, climb, cruise, and rapid descent. The multi-parameter calibration condition map is a high-dimensional knowledge graph that integrates parameter coupling topology networks and flight condition simulation distribution, realizing a unified expression of physical associations and task semantics. The calibration response path refers to the directed path connecting various flight condition segments in the multi-parameter calibration condition map. It defines the temporal logic of calibration execution, parameter weight configuration, and resource scheduling strategy. Its path planning needs to adapt to the dynamic characteristics of typical flight conditions such as hovering, climb, cruise, and rapid descent.

[0016] Typical flight conditions of a UAV refer to the representative steady-state and dynamic operating states of the UAV during mission execution, including hovering, climbing, cruise, and rapid descent. Each condition has distinguishable physical characteristics and parameter coupling modes. Priority parameters refer to dynamic weighting factors that reflect calibration requirements, including three core indicators: sensor linear margin, target calibration uncertainty, and measurement value reproducibility efficiency. Furthermore, sensor linear margin refers to the safety margin between the sensor's current operating point and the upper limit of its range, which is usually required to be >20%. Target calibration uncertainty refers to the extended uncertainty limit set according to airworthiness requirements, such as thrust measurement U≤0.5%, k=2. Measurement value reproducibility efficiency refers to the ability to complete calibration and recover the test within a unit of time, usually measured in minutes / parameter. Multi-parameter collaborative calibration optimization is a scalar indicator that quantitatively evaluates the comprehensive performance of calibration strategies under specific operating conditions. It comprehensively reflects the parameter coupling suppression capability, calibration accuracy achievement, and execution efficiency. The higher the score, the better the calibration configuration.

[0017] Execution steps: First, preprocess and detect anomalies in the massive calibration data. For example, using the Isolation Forest algorithm or the 3σ criterion, filter out a subset of anomalous data requiring intervention, such as a batch of tests where the thrust-current relationship deviates from the theoretical curve by more than 15%, to avoid invalid or misleading data interfering with subsequent modeling. Then, based on real flight telemetry data, construct a flight condition simulation distribution covering typical flight states. Simultaneously, use system identification or causal inference methods to establish a parameter coupling disturbance topology, revealing chain disturbance paths such as "motor temperature rise → increased winding resistance → current fluctuation → thrust attenuation". Finally, merge the two to generate a multi-parameter calibration condition map. This map identifies high-risk calibration areas such as the high-temperature + high-speed coupling zone, providing a semantic coordinate system for subsequent path planning. Preferably, static, discrete calibration points are upgraded to a dynamic, semantically related condition space, laying the data and structural foundation for task-driven collaborative calibration.

[0018] S3: Based on the calibration response path and the multi-parameter collaborative calibration optimization scores under each working condition section, perform multi-objective optimization and balance of the calibration strategy to obtain multi-parameter collaborative calibration instructions; S4: Use the multi-parameter collaborative calibration instructions to perform synchronous calibration and value correction of various test parameters during the test of the UAV power system.

[0019] Specifically, the multi-objective optimization balance of the calibration strategy refers to the process of seeking the Pareto optimal solution set among multiple optimization objectives. The trade-offs between different objectives are coordinated through a multi-objective evolutionary algorithm to avoid the extreme tendencies caused by single-objective optimization. The multiple optimization objectives include calibration accuracy, cycle coverage, and airworthiness compliance. The multi-parameter collaborative calibration command is a set of executable control commands output by the optimization process. It includes the calibration coefficient matrix of each test parameter, sampling timing configuration, filter parameter settings, and operating condition switching logic, which directly drives the test platform hardware to perform calibration operations.

[0020] Synchronous calibration refers to the parallel calibration operation performed on multiple test parameters such as thrust, speed, current, and temperature within the same time window. This differs from the serial mode of traditional sequential calibration and emphasizes the temporal consistency and coupling coordination among multiple parameters. Measurement correction refers to the process of compensating for errors in the original sensor measurements according to the correction coefficients in the calibration instructions. This includes operations such as zero-point drift correction, sensitivity coefficient adjustment, nonlinear error compensation, and cross-coupling decoupling. The testing process refers to the entire performance verification process of the UAV power system on the test platform, covering stages such as no-load self-test, static calibration, dynamic response testing, temperature rise assessment, and durability testing, with a duration ranging from several minutes to hundreds of hours.

[0021] Execution steps: Using the calibration response path as the time-operational axis, the optimization scores of each operating condition segment are used as the objective function input to construct a multi-objective optimization model. For example, in the hovering segment, the priority parameters may focus on thrust-current linearity and temperature drift stability, with a weight of 0.6; in the rapid descent segment, more attention is paid to dynamic response delay and vibration disturbance suppression, with a weight adjusted to 0.7. The Pareto optimal frontier is solved within 100 iterations using the NSGA-II algorithm (Non-dominated Sorting Genetic Algorithm II). Finally, the solution that satisfies the airworthiness constraints and has the highest comprehensive score is selected as the multi-parameter collaborative calibration command. The multi-parameter collaborative calibration command is usually sent to the real-time controller of the test platform in JSON or binary protocol form, triggering dynamic correction during the operation of the power system. For example, when the motor speed reaches 8000 rpm and the shell temperature rises to 65°C, the zero-point offset of the thrust sensor, the current sampling gain, and the temperature compensation coefficient are adjusted synchronously. Preferably, discrete, static calibration behavior is transformed into task semantic-driven closed-loop collaborative control, which significantly enhances the ability of test data to characterize real flight performance.

[0022] Furthermore, the method for filtering multi-parameter calibration datasets in this application includes: Receive power system monitoring data including multi-sensor channel identifiers, test bench equipment numbers, and test bench operating mode codes; perform alignment processing based on the power system monitoring data to obtain a multi-parameter calibration data set.

[0023] Specifically, the multi-sensor channel identifier is a coded identifier used to uniquely identify the measurement channels of different physical quantities in the UAV power system. It typically includes sensor type code, channel number, and range level information, and is stored in the data acquisition system as a 16-bit or 32-bit integer code. Furthermore, the sensor type code includes F for thrust sensor, R for speed encoder, I for current sensor, and T for temperature sensor. The test bench equipment number is a unique identifier assigned to each UAV power system test platform, using a string or barcode format, to distinguish the data source in a multi-test bench parallel testing scenario, and to achieve equipment-level data traceability and status management.

[0024] The test bench operating mode code is an coded signal representing the current operating status of the test bench. It is generated by the test bench control system based on sensor self-test status, environmental monitoring data, and load simulator feedback. It is usually an 8-bit or 16-bit binary code, with each segment corresponding to standby, no-load calibration, load test, dynamic response test, temperature rise assessment, and other modes and abnormal status indicators. The power system monitoring data is the raw physical quantity data stream collected in real time by each sensor of the test bench. It includes metadata such as timestamp, device number, channel identifier, quantity value, and operating mode code. The data update frequency is usually between 10Hz and 10kHz, depending on the sensor type and test requirements. Alignment processing refers to the process of eliminating differences in time dimension, format specifications, and semantic expression of multi-source heterogeneous data to achieve unified data organization. This includes operations such as timestamp synchronization, data format conversion, invalid data removal, and structured storage.

[0025] Execution steps: The system receives raw monitoring data streams from the test platform's data acquisition system. Each record carries a channel identifier, device number, and operating mode code. For example, in a climb test, data from the thrust sensor, motor speed encoder, ESC current, and housing temperature are received simultaneously. Subsequently, due to the different sampling frequencies of each sensor, alignment processing is required: using a high-precision clock as a reference, low-frequency signals are upsampled using linear interpolation or spline fitting. The operating mode code is used as an event anchor point to ensure strict alignment of all parameters during critical stages such as climb start, steady-state climb, and climb end, forming a structured multi-parameter calibration data set. Each row corresponds to a time point or operating condition and contains a complete parameter vector. Through these steps, a high-quality, spatiotemporally consistent input data foundation is constructed, avoiding false coupling or calibration deviations caused by asynchronous sampling.

[0026] Furthermore, based on the power system monitoring data, an alignment process is performed to obtain a multi-parameter calibration data set. The method of this application includes: The data is sorted by the timestamps corresponding to the acquisition of the power system monitoring data, and the test bench equipment number is used as the identification index. The data format is adapted by the multi-sensor channel identifier, and the format is standardized by a data mapping table. At the same time, the test bench working mode code is parsed, and then aligned with the identification index to obtain the multi-parameter calibration data set.

[0027] Specifically, the acquisition timestamp is an absolute time stamp generated by the sensor data acquisition card or distributed clock system. It typically uses the Unix timestamp precise time protocol format with an accuracy down to the microsecond level. It is used to characterize the exact moment of data sampling and serves as the benchmark for time-series alignment of multi-source data. The identifier index is a database index structure built using the test bench equipment number as the key value. It is implemented using a B+ tree or hash table and supports fast retrieval and grouping aggregation, organizing scattered data records into logical datasets according to equipment affiliation. The data format adaptation for multi-sensor channel identifiers refers to the process of automatically matching the corresponding data parsing rules and conversion algorithms based on the sensor type indicated by the multi-sensor channel identifier. This involves the identification and configuration of parameters such as measurement range, resolution, sampling rate, and physical units.

[0028] The data mapping table is a predefined channel configuration database that stores metadata information for each sensor channel, including channel identifier, sensor model, upper and lower range limits, sensitivity coefficient, output signal type, standard engineering units, and calibration history. It is usually stored locally or in the cloud in JSON or XML format. The output signal type includes voltage, current, and frequency. Format standardization is the process of converting the raw output signals of different sensors into standard engineering units, including linear conversion of analog signal quantization, counting / period conversion of pulse signals, frame parsing of digital protocols, and unit conversion. Parsing the test bench operating mode code further refers to the bitwise decoding of the test bench operating mode code, identifying the Boolean values ​​of each status flag bit, and determining the current operating mode and abnormal status of the test bench.

[0029] Execution steps: The raw monitoring data is globally sorted based on the high-precision acquisition timestamp to eliminate out-of-order issues caused by network latency or asynchronous acquisition. Data belonging to the same tested object is aggregated into independent data buckets using the test bench device number as an index to avoid data mixing from multiple devices. Next, the raw messages corresponding to the multi-sensor channel identifiers are parsed and their units are unified through a preset data mapping table. For example, the 12-bit ADC value of a Hall current sensor is converted to a standard ampere value using a scaling factor of 0.0488A / bit, and missing fields are filled with NaN.

[0030] Simultaneously, the test bench's working mode code is parsed using a lookup table to identify the representative rapid descent condition and obtain the start and end time windows of that condition. Under each device index, based on the parsed working condition time period, all standardized sensor data are aligned using a sliding window or resampled based on event triggering to form a multi-parameter calibration data set segmented by working condition, arranged by time sequence, and aligned by channel. Through the above steps, the structure and semantic integrity of the data are improved.

[0031] Furthermore, the method for parsing the working mode code of the test bench includes: The test bench operating mode code is decoded according to the preset mode encoding rules to identify the test bench operating status; if the test bench operating status is standby or non-testing, the corresponding power system monitoring data is determined to be invalid and is discarded.

[0032] Specifically, the preset mode coding rule is a pre-agreed binary coding protocol between the test bench control system and the data processing system. It defines the bit segment structure, status flag bit allocation, and verification mechanism of the test bench working mode code. It usually adopts a 16-bit or 32-bit fixed-length code. The high 4 bits represent the main operating mode, the middle 8 bits represent the sub-status flags, and the low 4 bits are used as parity check or CRC check bits to ensure the reliability of the code transmission. The high 4 bits further represent the main operating mode, such as standby, no-load calibration, load test, dynamic response test, and temperature rise assessment; the middle 8 bits represent the sub-status flags, such as sensor self-test completed, ambient temperature and humidity normal, load analog signal valid, and communication link normal.

[0033] Decoding according to preset mode encoding rules refers to the process of parsing the original binary code by bit operations according to preset mode encoding rules and extracting the semantic information of each field, involving shift operations, masking operations and table lookup matching; the test bench operating status is a comprehensive representation of the test bench's working condition at a specific moment, determined by the logical combination of the main operating mode and sub-state flags, reflecting whether the test bench is in an effective data acquisition condition; the standby state refers to the idle state after the test bench is powered on and initialized, but before receiving test task instructions. At this time, the sensor is powered on but no load is applied, and the data output is a static zero-point value, which has no calibration or testing value.

[0034] Non-test state refers to an abnormal state in which the test bench receives task instructions but cannot enter effective testing due to unmet preconditions. This includes situations such as sensors not completing self-tests, ambient temperature and humidity exceeding the calibration allowable range, and missing load analog signals. Further, a sensor not completing self-tests is indicated by a sub-status flag of 0; ambient temperature and humidity exceeding the calibration allowable range are indicated by temperatures >40℃ or <5℃, and relative humidity >85%RH; and missing load analog signals are indicated by the eddy current brake not responding. Determining the corresponding power system monitoring data to be invalid refers to the decision-making process of performing Boolean logic judgments based on the parsed status flags to generate data validity labels. Data validity labels include valid labels and invalid labels. Removal refers to the data cleaning operation of removing or isolating data records marked as invalid from the multi-parameter calibration data set to prevent them from participating in subsequent calibration calculations.

[0035] Execution steps: Based on preset mode encoding rules, the test bench operating mode code attached to each power system monitoring data is decoded in real time. For example, when mode code 0x00 is received, it is identified as a standby state; if it is 0x05, it may correspond to a non-test state where the motor self-test has been completed but no load has been applied. Once the operating state is identified as standby or non-test, the data validity judgment logic is triggered, marking all associated multi-sensor data within that time period as invalid. Although these data have physical authenticity, they lack the coupling response characteristics of the power system under real load. If mixed into the calibration data set, they will dilute the effective operating condition density and even interfere with the anomaly detection model, such as misjudging normal standby current as low load deviation. Therefore, the system automatically removes them from the data stream. Preferably, after this screening, the proportion of effective calibration data is increased, and the signal-to-noise ratio of multi-parameter correlation analysis is improved, ensuring the purity and task relevance of the data input to the calibration model.

[0036] Furthermore, the method of this application includes: The test bench's operating states include standby state, no-load calibration state, load test state, dynamic response test state, and temperature rise assessment state; among which, the non-test states include states where the sensor is not powered on and has completed its self-test, the ambient temperature and humidity exceed the allowable calibration range, and the load simulation signal is missing.

[0037] Specifically, the no-load calibration state refers to the special state in which the test bench performs zero-point calibration and sensitivity verification of the sensor system without applying any external mechanical load. At this time, the test bench only drives the motor to run idle or stand still, and simulates the sensor output through standard signal sources such as precision resistance boxes and frequency generators to verify the linearity and repeatability of the acquisition channel. This usually lasts for 5-15 minutes and is a basic step to ensure the accuracy of subsequent load test data. The test bench does not apply any external mechanical load, that is, the propeller or motor of the power system has no aerodynamic / mechanical load.

[0038] Load testing refers to the steady-state test state in which a controllable mechanical load is applied to the power system by a test bench using load simulation devices such as magnetic powder brakes, eddy current dynamometers, or hydraulic dynamometers to simulate the aerodynamic load of a propeller. In this state, parameters such as thrust, speed, current, and temperature reach dynamic equilibrium. This is used to measure the steady-state performance curves of the power system, such as thrust-power curves and efficiency-speed curves. A single test lasts from several minutes to tens of minutes. Dynamic response testing refers to the transient test state in which the test bench performs rapid load changes or frequency scans. This is used to evaluate the response speed, stability, and control bandwidth of the power system. The data acquisition frequency usually needs to be increased to 1-10kHz to capture the transient process. Rapid load changes include stepping from 50% load to 70% load.

[0039] Temperature rise assessment refers to the durability test state in which the test bench operates continuously under rated load or overload conditions, monitoring the temperature changes of components such as motors, ESCs, and batteries, and verifying the effectiveness of thermal design margins and overheat protection mechanisms. Sensor not powered on and self-test completed means that the sensor has not completed its internal diagnostic program after being powered on. At this time, the output data is unreliable and includes power-on impulse noise or default placeholder values. Ambient temperature and humidity exceeding the calibration allowable range means that the temperature or humidity of the environment where the test bench is located exceeds the operating specifications of the sensor and standard, such as the temperature exceeding (20±5)℃ or the relative humidity exceeding 85%RH, which leads to sensor sensitivity drift or condensation risk. Load analog signal missing means that the load device does not feed back a valid torque or speed signal, which may be caused by communication failure, driver alarm, or loose mechanical connection. At this time, closed-loop load control cannot be established.

[0040] Execution steps: After decoding the test bench's operating status, further refine the criteria for determining valid and invalid data, identifying five types of valid test states. For example, when the mode code corresponds to the dynamic response test state, the expected data should include high-frequency speed fluctuations and current spikes; if it is a temperature rise assessment state, a temperature rise curve of ≥30 minutes needs to be continuously recorded; conversely, once any non-test state characteristic is detected, such as the ambient temperature and humidity sensor reporting 42℃, exceeding the calibration allowable upper limit of 40℃; or the load torque signal remaining unchanged for 500ms, the current batch of data is determined to be invalid. In one test, due to an air conditioning malfunction, the laboratory humidity suddenly rose to 85%RH. Although the equipment displayed a load test state, the system automatically marked all thrust and current records for that period as invalid and removed them based on environmental monitoring data. This refined criterion can prevent false valid data from contaminating the calibration model. Preferably, it involves constructing a data reliability filtering mechanism oriented towards real calibration scenarios to ensure that multi-parameter collaborative calibration is based only on valid observation samples with clear physical meaning and complete load.

[0041] Furthermore, this application proposes a multi-parameter calibration condition map that includes parameter-coupled perturbation topology and flight condition simulation distribution. The method includes: Acquire typical flight profile data for UAV airworthiness certification, extract hovering, climb, cruise, and rapid descent nodes and parameter association rules, and configure a parameter coupling topology network; collect historical calibration data, identify the measurement deviation amplitude and cross sensitivity of various test parameters under different operating conditions; based on the parameter coupling topology network, superimpose the measurement deviation amplitude and cross sensitivity to obtain the multi-parameter calibration operating condition map.

[0042] Specifically, typical flight profile data for UAV airworthiness certification refers to the standard flight mission profile of UAVs approved by the reviewing party during the type approval or production license approval process, in accordance with airworthiness requirements such as civil aviation regulations. It includes time-series data of longitude, latitude, altitude, speed, and attitude angle, covering the complete flight phases such as takeoff, climb, cruise, maneuver, descent, and landing. It is usually stored in CSV or MAT format with a sampling frequency of 1-10Hz. The hovering condition node refers to the steady-state flight segment in the flight profile where the vertical velocity and airspeed are both zero and the altitude remains stable. At this time, the power system is at the high-power hovering efficiency point, the thrust is equal to the gravity, and the speed and current are maintained at a high level of balance. This is the typical operating state of logistics UAVs and inspection UAVs.

[0043] The climb condition refers to the dynamic flight segment where the vertical velocity is greater than 0 and the airspeed gradually increases. The power system needs to output excess thrust to overcome gravity and drag. The transient peak current can reach 150%-200% of the rated value, and the motor temperature rise rate is the fastest. It is a key assessment point for thermal design. The cruise condition refers to the steady-state flight segment where the vertical velocity is zero and the airspeed maintains the economic cruise speed. The power system is in the optimal efficiency range with minimal parameter fluctuations. It is the core condition for evaluating endurance performance. The descent condition refers to a special flight segment where the vertical velocity is less than 0 and the absolute value is large. It may be accompanied by negative propeller torque. The power system enters a regenerative braking or deceleration state. The current direction may reverse, and the control stability risk is the highest.

[0044] The parameter association rules are a knowledge base describing the physical constraint relationships between thrust, speed, current, and temperature under various operating conditions, including theoretical formulas and empirical thresholds. The parameter coupling topology network is a graph structure model with test parameters as nodes and physical association strength as edges. The edge weights are determined by theoretical derivation or data-driven methods, representing the energy flow, signal flow, and disturbance transmission paths between parameters. The historical calibration data is a time-series database of original sensor values, standard reference values, environmental conditions, and correction coefficients accumulated by the test bench in past operating cycles, with a data volume reaching TB level.

[0045] Measurement deviation amplitude refers to the systematic deviation between the sensor reading and the standard reference value under a specific operating condition range. It is expressed as a percentage of full scale or absolute error and reflects the accuracy level of the sensor. Cross sensitivity refers to the coupling coefficient that causes a change in the measured value of one parameter to cause a change in the measured value of another parameter. For example, a 1A change in current causes a 0.05N temperature drift in the thrust sensor, or a 10°C increase in temperature causes a 0.3% deviation in the speed measurement. Furthermore, by superimposing the measurement deviation amplitude and cross sensitivity, the structural knowledge of the topology network and the statistical knowledge of deviation sensitivity are fused and mapped to generate a high-dimensional knowledge expression that combines physical interpretability and data adaptability.

[0046] Execution steps: Extract typical operational condition nodes from flight mission profiles that meet airworthiness certification requirements. For example, in a 30-minute mission profile that conforms to industry consensus, identify 4 hovering nodes, 2 5° climb segments, 1 60km / h cruise segment, and 1 simulated obstacle avoidance descent. Subsequently, based on massive historical calibration records, analyze the systematic deviations of various parameters under each operational condition. For example, in the climb condition, the average thrust is too high, and in the descent phase, the current sensor exhibits random fluctuations due to vibration interference. Simultaneously, calculate the cross-sensitivity moments through multiple regression or sensitivity analysis. The algorithm identifies the cross-sensitivity of temperature to current. These quantified results are then injected into a parameter-coupled topology network: using a thrust-current-temperature triangular subgraph, with edge weights set as the deviation correlation coefficient and cross-sensitivity, respectively, this enhanced topology network is mapped onto a flight profile coordinate system consisting of hovering, climb, cruise, and rapid descent nodes, forming a multi-parameter calibration condition map. This map not only marks the dominant deviation sources in each condition segment but also uses color gradients to display the intensity of coupled disturbances, enabling the calibration strategy to accurately focus on high-risk areas. Ideally, this achieves the construction of a calibration knowledge graph integrating general calibration, task, error, and coupling, serving as the core knowledge foundation supporting the generation of collaborative calibration instructions.

[0047] Furthermore, by combining priority parameters, the multi-parameter collaborative calibration optimization score is obtained for each operating condition section. The method of this application also includes: According to the calibration response path, the parameter coupling strength marked under each working condition segment in the multi-parameter calibration working condition map is called; the corresponding target calibration uncertainty and value reproduction efficiency are confirmed by the parameter coupling strength marked under each working condition segment; the sensor linear margin, target calibration uncertainty, and value reproduction efficiency are used as calibration evaluation indicators, and the priority parameters are set.

[0048] Specifically, parameter coupling strength is a scalar index that quantifies the interaction strength between various test parameters under a specific operating condition. It is calculated by weighting the edge weights of the parameter coupling topology network with the cross-sensitivity of historical data, and its value range is usually [0,1]. 0 indicates that the parameters are completely independent, and 1 indicates strong coupling, such as the decisive influence of current change on temperature measurement. It is stored as an edge attribute in the multi-parameter calibration operating condition map. Target calibration uncertainty is the upper limit of the expanded uncertainty of the calibration measurement result set for a specific operating condition based on the accuracy requirements of the test task and the performance specifications of the sensor. For example, U≤0.3% for thrust measurement under hovering condition and U≤0.8% for current measurement under climbing condition. It is the core evaluation index of calibration quality.

[0049] Measurement reproducibility efficiency refers to the average time required to complete calibration and restore to a test-ready state within a specific operating condition range. This includes the entire process time from standard access, data acquisition, algorithm calculation, correction application, and verification. It is usually measured in minutes per parameter or seconds per operating condition range. Sensor linear margin refers to the safe margin ratio between the current operating point of the sensor and its upper limit of range. The calculation formula is (upper limit of range - current operating point) / upper limit of range × 100%. It is usually required to be >20% to avoid the risk of nonlinear saturation. It is a dynamic indicator for evaluating the health status of the sensor and the feasibility of calibration.

[0050] The calibration evaluation index is a set of quantitative dimensions used to comprehensively evaluate the merits of calibration strategies. It is established as a three-dimensional index system: linear margin, target calibration uncertainty, and value reproduction efficiency. Among them, linear margin reflects the sensor status, target calibration uncertainty reflects the accuracy requirement, and value reproduction efficiency reflects the time constraint. The priority parameter is a guiding factor that dynamically adjusts the weights of the above three-dimensional indexes according to the current test scenario. Through normalization and weighted combination, it forms a single-valued or multi-valued optimization target vector, which drives the search direction of subsequent multi-objective optimization algorithms.

[0051] Execution steps: Using the calibration response path as a navigation clue, query the parameter coupling strength of the corresponding section in the multi-parameter calibration condition map segment by segment. For example, in the climb acceleration section, the map shows that the coupling strengths of thrust-current-temperature rise are 0.78, 0.65 and 0.59 respectively, indicating that there is a strong nonlinear interaction in this section. Based on this, dynamically set the target calibration uncertainty: the stronger the coupling, the more stringent the allowable uncertainty. For example, the uncertainty is set to U=0.8% in the climb section, while it can be relaxed to U=2.0% in the standby section.

[0052] Meanwhile, highly coupled sections typically require denser sampling or multiple iterations, thus adjusting the efficiency target for value reproduction accordingly. For example, the climbing section is allowed 30 seconds, while the hovering section is compressed to 18 seconds. Furthermore, the linear margin of each sensor is assessed based on its current health status. If the nonlinear error threshold is 0.5%, and the aging of a current sensor has resulted in a nonlinear error of 0.45%, then its linear margin is only 0.05%, requiring higher priority compensation during calibration. These three indicators are structured into a calibration evaluation vector, and priority parameters are set according to the task type. Preferably, static calibration indicators are transformed into dynamic optimization targets that are task-adaptive and condition-aware, providing accurate and quantifiable decision-making basis for multi-objective balancing.

[0053] Furthermore, based on the calibration response path and the multi-parameter collaborative calibration optimization scores under each operating condition segment, a multi-objective optimization balance of the calibration strategy is performed to obtain a multi-parameter collaborative calibration instruction. The method of this application includes: Based on a multi-objective function, a local search strategy is introduced in the crossover and mutation operation of the initial calibration parameter population. This local search strategy is used to fine-tune the new calibration parameter combinations and, combined with a crowding mechanism, to screen non-dominated solutions in the Pareto optimal solution set to determine the multi-parameter collaborative calibration instruction.

[0054] Specifically, the multi-objective function is a mathematical expression for simultaneously optimizing multiple conflicting objectives. It is a function vector of three objectives—multi-parameter coupled calibration accuracy, calibration cycle coverage, and airworthiness compliance achievement rate—in the form F(x)=(f1(x),f2(x),f3(x)), where x is a combination of calibration parameters. There is a competitive relationship between the objectives, with f1(x) corresponding to multi-parameter coupled calibration accuracy, f2(x) corresponding to quasi-cycle coverage, and f3(x) corresponding to airworthiness compliance achievement rate. The initial calibration parameter population is the initial solution set of the multi-objective optimization algorithm, consisting of multiple sets of individual calibration parameters. Each individual encodes the calibration configuration for a specific operating condition section, including filter coefficients, sampling rate, and correction matrix elements. The initialization method can be random generation, Latin hypercube sampling, or empirical heuristics based on historical best solutions.

[0055] Crossover is an evolutionary operator in genetic algorithms that simulates biological gene recombination. It generates new individuals by exchanging partial gene segments between two parent individuals through single-point crossover, uniform crossover, etc. The crossover probability Pc is usually set to 0.8-0.95 and is used for global search to explore new solution space regions. Mutation is a random perturbation mechanism that simulates gene mutation. It randomly changes certain gene positions of an individual with a small probability Pm, such as Gaussian noise or bit flip, to maintain population diversity and prevent premature convergence. Local search strategy is a fine-grained optimization mechanism embedded in the global evolutionary framework. It performs small-step searches in the neighborhood of the new individuals generated by crossover and mutation using gradient descent, pattern search, simulated annealing, etc., to accelerate convergence to local optima and balance global exploration with local development.

[0056] Fine-tuning refers to the operation of making small adjustments to calibration parameters during local search. The adjustment range is usually limited to ±5% to ±10% of the parent value to avoid disrupting the existing good structure. The crowding mechanism is a selection strategy in the NSGA-II algorithm to maintain the diversity of Pareto front solution distribution. By calculating the density distance of solutions in the target space, i.e., the sum of the Euclidean distances of adjacent solutions, solutions with large crowding distances are preferred to prevent the solution set from clustering in local regions. A non-dominated solution is a solution in multi-objective optimization where no other solution is better than it on all objectives; that is, for solution x, there is no y such that f (y)≤f (x) For all i, at least one strict inequality holds, and these solutions constitute the Pareto optimal solution set; non-dominated solutions in the Pareto optimal solution set are screened, and further, from the candidate solution set generated by evolution, the next generation population or the final solution is selected based on the non-dominated ranking level and the crowding distance.

[0057] Execution steps: Under the multi-objective optimization framework, an objective function vector is defined. The initial population consists of 200 sets of calibration parameter combinations, each containing correction coefficients for four types of sensors: thrust, current, speed, and temperature. In each generation of evolution, after performing SBX crossover (Simulated Binary Crossover) and polynomial mutation to generate offspring, a local search is immediately applied to each new individual: for example, the gradient of the objective function is estimated using the finite difference method, and 5 steps of Nelder-Mead (simplex optimization method) are performed within ±2% of the parameter space to make the solution closer to the local Pareto front. Subsequently, the parent and offspring generations are merged, and the first front solution set is selected based on non-dominated sorting. Crowding calculation based on Euclidean distance is introduced. If the distance between a solution and its nearest neighbor solution in the objective space is less than a threshold, it is considered crowded, reducing its probability of being selected into the next generation. After multiple generations of iteration, a uniformly distributed and comprehensive Pareto optimal solution set is obtained, usually containing 30–50 non-dominated solutions. The best-matching solution is selected from this set according to the current task type and converted into structured instructions. Preferably, the introduction of local search and congestion mechanisms to filter feasible solutions is the core optimization engine for generating robust, task-adaptive multi-parameter collaborative calibration instructions.

[0058] Furthermore, the method of this application includes: A multi-objective function is constructed, which includes multi-parameter coupled calibration accuracy, calibration cycle coverage, and airworthiness compliance achievement rate. At the same time, based on the calibration response path, the initial calibration parameter population is encoded based on the multi-parameter calibration condition map, so that each set of initial calibration parameters corresponds to the calibration requirements of each condition segment.

[0059] Specifically, multi-parameter coupled calibration accuracy is an indicator that comprehensively evaluates the accuracy of measurement results after multi-parameter collaborative calibration. Unlike traditional single-parameter accuracy evaluation, this indicator explicitly considers the correction effect of coupling effects between parameters, comprehensively reflecting the joint measurement accuracy of parameters such as thrust, speed, current, and temperature under typical operating conditions. Calibration cycle coverage rate is an efficiency indicator that measures the proportion of operating condition segments covered within a unit calibration cycle. It is defined as the ratio of the actual number of covered operating condition segments to the total number of operating condition segments in the map, or the ratio of calibration execution time to the available time window, reflecting the time economy and task completeness of the calibration process. Airworthiness compliance achievement rate is an indicator that evaluates the degree of compliance of post-calibration test data with airworthiness regulations. It is usually measured by the proportion of data that meets the measurement uncertainty limits of key parameters of the power system in CCAR-92, and is a key threshold indicator for whether the calibration results can support type certification.

[0060] Encoding based on multi-parameter calibration condition maps refers to an encoding method that maps and associates the genotypic structure of individual calibration parameters with the topological structure of the condition map. This allows each set of gene fragments of an individual to correspond to the calibration configuration of a specific condition segment, forming a three-layer correspondence between condition segment, gene fragment, and calibration parameter. This differs from traditional random coding and possesses physical semantics and structural constraints. Calibration requirements refer to the functional requirements of a specific condition segment on the calibration strategy, including accuracy requirements, efficiency requirements, and stability requirements. These requirements are jointly determined by the parameter coupling strength, flight mission criticality, and sensor status of that segment.

[0061] Execution steps: Construct a multi-objective function F(x)=(f1(x),f2(x),f3(x)), where f1(x) corresponds to the multi-parameter coupled calibration accuracy, and f1(x) is positively correlated with the multi-parameter coupled calibration accuracy -1. Maximizing accuracy is equivalent to minimizing error. f2(x) corresponds to the quasi-cycle coverage rate, and f2(x) is positively correlated with the calibration cycle coverage rate. It is defined as the ratio of the total time of the operating condition segments effectively covered by the calibration command to the total duration of the mission profile. f3(x) corresponds to the airworthiness compliance achievement rate, and f3(x) is positively correlated with the airworthiness compliance achievement rate. Subsequently, in the initial calibration parameter population, each group of individuals is segmented and coded according to the operating condition segments traversed by the calibration response path to ensure that each operating condition segment has a dedicated calibration sub-parameter. This coding method enables the optimization process to allocate more refined correction resources for highly coupled segments such as rapid descent and simplify the processing of low-sensitivity segments such as standby. Preferably, a multi-objective function is constructed to provide an objective standard for evaluating the merits of the calibration strategy; the coding method based on the operating condition map embeds physical knowledge into the search space, so that the evolution process takes place in the subspace that explicitly corresponds to the operating condition segment, which significantly improves the convergence efficiency and makes the calibration instructions of the optimized output directly interpretable. The configuration of each segment can be traced back to its physical requirements, which is convenient for engineers to understand and debug.

[0062] In summary, the beneficial effects of the embodiments of this application are: By filtering the multi-parameter calibration dataset to identify abnormal test bench datasets requiring calibration intervention, and formulating a multi-parameter calibration condition map including parameter coupling disturbance topology and flight condition simulation distribution, this application provides a multi-parameter collaborative calibration method and apparatus for a UAV power system test platform. This method constructs a multi-parameter calibration condition map, sets calibration response paths adapted to typical UAV flight conditions, and, combined with priority parameters, obtains multi-parameter collaborative calibration optimization scores for each condition segment. Based on the calibration response paths and the multi-parameter collaborative calibration optimization scores for each condition segment, a multi-objective optimization balance of the calibration strategy is performed to obtain multi-parameter collaborative calibration instructions. Using these instructions, synchronous calibration and value correction of various test parameters are executed during UAV power system testing. This achieves the technical effect of constructing a multi-parameter calibration condition map, setting calibration response paths adapted to typical flight conditions, performing multi-objective optimization balance of the calibration strategy to generate executable collaborative calibration instructions, and improving multi-parameter collaborative consistency while maintaining calibration accuracy.

[0063] Example 2, based on the same inventive concept as the multi-parameter collaborative calibration method for the UAV power system test platform in the foregoing examples, such as... Figure 2 As shown in the embodiment of this application, a multi-parameter collaborative calibration device for an unmanned aerial vehicle (UAV) power system test platform is provided, wherein the device includes: Data filtering module M100: filters multi-parameter calibration data sets, identifies abnormal test bench datasets that require calibration intervention, and drafts multi-parameter calibration condition maps that include parameter coupling disturbance topology and flight condition simulation distribution.

[0064] Calibration response path setting module M200: Sets a calibration response path adapted to the typical flight conditions of the UAV in the multi-parameter calibration condition map, and obtains the multi-parameter collaborative calibration optimization score under each condition segment by combining the priority parameters.

[0065] The optimization balancing module M300 performs multi-objective optimization balancing of the calibration strategy based on the calibration response path and the multi-parameter collaborative calibration optimization scores under each operating condition segment, and obtains multi-parameter collaborative calibration instructions.

[0066] Synchronous calibration module M400: Using the multi-parameter collaborative calibration command, it performs synchronous calibration and value correction of various test parameters during the test of the UAV power system.

[0067] Furthermore, the data filtering module M100 is used to perform the following method: Receive power system monitoring data including multi-sensor channel identifiers, test bench equipment numbers, and test bench operating mode codes; perform alignment processing based on the power system monitoring data to obtain a multi-parameter calibration data set.

[0068] Furthermore, the data filtering module M100 is also used to perform the following method: The data is sorted by the timestamps corresponding to the acquisition of the power system monitoring data, and the test bench equipment number is used as the identification index. The data format is adapted by the multi-sensor channel identifier, and the format is standardized by a data mapping table. At the same time, the test bench working mode code is parsed, and then aligned with the identification index to obtain the multi-parameter calibration data set.

[0069] Furthermore, the data filtering module M100 is also used to perform the following method: The test bench operating mode code is decoded according to the preset mode encoding rules to identify the test bench operating status; if the test bench operating status is standby or non-testing, the corresponding power system monitoring data is determined to be invalid and is discarded.

[0070] Furthermore, the data filtering module M100 is also used to perform the following method: The test bench's operating states include standby state, no-load calibration state, load test state, dynamic response test state, and temperature rise assessment state; among which, the non-test states include states where the sensor is not powered on and has completed its self-test, the ambient temperature and humidity exceed the allowable calibration range, and the load simulation signal is missing.

[0071] Furthermore, the data filtering module M100 is also used to perform the following method: Acquire typical flight profile data for UAV airworthiness certification, extract hovering, climb, cruise, and rapid descent nodes and parameter association rules, and configure a parameter coupling topology network; collect historical calibration data, identify the measurement deviation amplitude and cross sensitivity of various test parameters under different operating conditions; based on the parameter coupling topology network, superimpose the measurement deviation amplitude and cross sensitivity to obtain the multi-parameter calibration operating condition map.

[0072] Furthermore, the calibration response path setting module M200 is also used to perform the following method: According to the calibration response path, the parameter coupling strength marked under each working condition segment in the multi-parameter calibration working condition map is called; the corresponding target calibration uncertainty and value reproduction efficiency are confirmed by the parameter coupling strength marked under each working condition segment; the sensor linear margin, target calibration uncertainty, and value reproduction efficiency are used as calibration evaluation indicators, and the priority parameters are set.

[0073] Furthermore, the optimization balancing module M300 is used to perform the following method: Based on a multi-objective function, a local search strategy is introduced in the crossover and mutation operation of the initial calibration parameter population. This local search strategy is used to fine-tune the new calibration parameter combinations and, combined with a crowding mechanism, to screen non-dominated solutions in the Pareto optimal solution set to determine the multi-parameter collaborative calibration instruction.

[0074] Furthermore, the optimization balancing module M300 is also used to perform the following methods: A multi-objective function is constructed, which includes multi-parameter coupled calibration accuracy, calibration cycle coverage, and airworthiness compliance achievement rate. At the same time, based on the calibration response path, the initial calibration parameter population is encoded based on the multi-parameter calibration condition map, so that each set of initial calibration parameters corresponds to the calibration requirements of each condition segment.

[0075] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Figure 1 The multi-parameter collaborative calibration method and specific examples of the UAV power system test platform in Embodiment 1 are also applicable to the multi-parameter collaborative calibration device of the UAV power system test platform in this embodiment. Through the foregoing detailed description of the multi-parameter collaborative calibration method of the UAV power system test platform, those skilled in the art can clearly understand the multi-parameter collaborative calibration device of the UAV power system test platform in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0076] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0077] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A multi-parameter collaborative calibration method for an unmanned aerial vehicle (UAV) power system test platform, characterized in that, The method includes: The multi-parameter calibration dataset was screened to identify the abnormal test bench datasets that required calibration intervention, and a multi-parameter calibration condition map including parameter coupling disturbance topology and flight condition simulation distribution was drafted. In the multi-parameter calibration condition map, a calibration response path adapted to the typical flight conditions of the UAV is set, and combined with the priority parameters, the multi-parameter collaborative calibration optimization score under each condition segment is obtained. Based on the calibration response path and the multi-parameter collaborative calibration optimization scores under each operating condition section, the calibration strategy is optimized and balanced to obtain the multi-parameter collaborative calibration instruction. The multi-parameter collaborative calibration command is used to perform synchronous calibration and value correction of various test parameters during the testing of the UAV power system.

2. The multi-parameter collaborative calibration method for the UAV power system test platform as described in claim 1, characterized in that, The method for filtering multi-parameter calibration datasets includes: Receive power system monitoring data including multi-sensor channel identifiers, test bench equipment numbers, and test bench operating mode codes; Based on the power system monitoring data, an alignment process is performed to obtain a multi-parameter calibration data set.

3. The multi-parameter collaborative calibration method for the UAV power system test platform as described in claim 2, characterized in that, Based on the power system monitoring data, an alignment process is performed to obtain a multi-parameter calibration data set. The method includes: The data is sorted by the timestamp of the power system monitoring data and the test bench equipment number is used as the identifier index. The data format adapted by the multi-sensor channel identifier is standardized using a data mapping table. Simultaneously, the test bench working mode code is parsed, and then aligned with the identifier index to obtain the multi-parameter calibration data set.

4. The multi-parameter collaborative calibration method for the UAV power system test platform as described in claim 3, characterized in that, The method for parsing the working mode code of the test bench includes: The test bench operating mode code is decoded according to a preset mode encoding rule to identify the test bench operating status. If the test bench is in standby or non-testing state, the corresponding power system monitoring data is determined to be invalid and discarded.

5. The multi-parameter collaborative calibration method for the UAV power system test platform as described in claim 4, characterized in that, The test bench's operating states include standby state, no-load calibration state, load test state, dynamic response test state, and temperature rise assessment state. The non-test states include states where the sensor is not powered on and has not completed its self-test, where the ambient temperature and humidity exceed the allowable calibration range, and where the load analog signal is missing.

6. The multi-parameter collaborative calibration method for the UAV power system test platform as described in claim 1, characterized in that, The method for constructing a multi-parameter calibration condition map that includes parameter-coupled perturbation topology and flight condition simulation distribution includes: Obtain typical flight profile data for UAV airworthiness certification, extract hovering, climb, cruise, and rapid descent nodes and parameter association rules, and configure parameter coupling topology network. Collect historical calibration data to identify the measurement deviation amplitude and cross sensitivity of various test parameters under different operating conditions; Based on the parameter-coupled topology network, the multi-parameter calibration condition spectrum is obtained by superimposing the measurement deviation amplitude and cross sensitivity.

7. The multi-parameter collaborative calibration method for the UAV power system test platform as described in claim 6, characterized in that, The method further includes: combining priority parameters to obtain multi-parameter collaborative calibration optimization scores for each operating condition segment; and obtaining multi-parameter collaborative calibration optimization scores for each operating condition segment. Based on the calibration response path, the parameter coupling strength marked under each working condition section in the multi-parameter calibration working condition map is called; By identifying the parameter coupling strength under each working condition section, the corresponding target calibration uncertainty and value reproducibility efficiency are confirmed. Sensor linear margin, target calibration uncertainty, and value reproducibility efficiency are used as calibration evaluation indicators, and the aforementioned priority parameters are set.

8. The multi-parameter collaborative calibration method for the UAV power system test platform as described in claim 7, characterized in that, Based on the calibration response path and the multi-parameter collaborative calibration optimization scores under each operating condition segment, a multi-objective optimization balance of the calibration strategy is performed to obtain a multi-parameter collaborative calibration instruction. The method includes: Based on a multi-objective function, a local search strategy is introduced in the crossover and mutation operation of the initial calibration parameter population. This local search strategy is used to fine-tune the new calibration parameter combinations and, combined with a crowding mechanism, to screen non-dominated solutions in the Pareto optimal solution set to determine the multi-parameter collaborative calibration instruction.

9. The multi-parameter collaborative calibration method for the UAV power system test platform as described in claim 8, characterized in that, The method includes: Construct a multi-objective function that includes multi-parameter coupled calibration accuracy, calibration cycle coverage, and airworthiness compliance achievement rate; Simultaneously, based on the calibration response path, the initial calibration parameter population is encoded based on the multi-parameter calibration condition map, so that each set of initial calibration parameters corresponds to the calibration requirements of each operating condition segment.

10. A multi-parameter collaborative calibration device for an unmanned aerial vehicle (UAV) power system test platform, characterized in that, The apparatus for implementing the multi-parameter collaborative calibration method for the unmanned aerial vehicle power system test platform according to any one of claims 1-9, wherein the apparatus comprises: Data filtering module: Filters multi-parameter calibration datasets, identifies abnormal test bench datasets requiring calibration intervention, and drafts multi-parameter calibration condition maps that include parameter coupling disturbance topology and flight condition simulation distribution; Calibration response path setting module: Sets a calibration response path adapted to the typical flight conditions of the UAV in the multi-parameter calibration condition map, and obtains the multi-parameter collaborative calibration optimization score under each condition segment by combining the priority parameters. The optimization and balancing module performs multi-objective optimization and balancing of calibration strategies based on the calibration response path and the multi-parameter collaborative calibration optimization scores under each operating condition segment, and obtains multi-parameter collaborative calibration instructions. Synchronous calibration module: Using the multi-parameter collaborative calibration command, it performs synchronous calibration and value correction of various test parameters during the test of the UAV power system.

Images